Designing sustainable landscapes that conserve biodiversity and produce food, water and energy is a global priority (Williams et al., 2017). Implementing sustainable landscapes requires knowledge of the key drivers of habitat loss and fragmentation in productive landscapes (Laurance et al., 2014). Landscapes embedded in a matrix of anthropogenic activities contain remnants of natural habitats of different sizes and shapes. Thus, landscapes with an intermediate level of disturbance have been shown to have greater spatial heterogeneity promoting biodiversity conservation in productive landscapes (Arroyo- Rodríguez et al., 2020). Therefore, an assessment of species responses is essential to identify optimal scenarios for biodiversity in productive landscapes with higher levels of disturbance (Arroyo- Rodríguez et al., 2020).

Three hypotheses have emerged to explain changes in biodiversity in anthropized landscapes: (i) The habitat amount hypothesis (see Fahrig, 2013) posits that the sample area effect overrides spatial habitat configuration and predicts that the amount of habitat in a local landscape, regardless of configuration, is the main landscape-level driver of species diversity. (ii) The spatial arrangement of habitat patches in human-dominated landscapes suggests that agricultural practices that promote different spatial arrangements of landscapes (e.g., greater structural complexity of landscape form) are often associated with improved habitat conditions for many species and that fragmentation and isolation of natural habitats threaten biodiversity beyond habitat loss (Fletcher et al., 2018). (iii) Habitat heterogeneity suggests that diverse habitats with complex spatial patterns support more species and functions (Martin et al., 2020). However, the evidence for these three main drivers of biodiversity in productive landscapes (hereafter agroscapes) has mainly focused on species richness, but not on other dimensions of biodiversity such as changes in species composition (β-diversity) or species diversity, dominance or the habitat requirements of species.

Ecological traits, which can be specific (e.g., body size, thermoregulatory capacity, or habitat specificity), determine the ability of species to adapt to anthropogenic landscapes (Tonetti et al., 2023). For example, forest specialist species may be more sensitive to landscape-scale habitat availability than habitat generalists (Sekercioglu, 2012). The decline of a particular ecological group can lead to changes in ecosystem functioning (Lu et al., 2024). In contrast, recent research has shown that changes in the composition and configuration of productive landscapes promote changes in alpha (local diversity) and beta diversity (changes in species composition) (Leite et al., 2022). Therefore, it is important to understand how species that perform different ecological functions are affected by habitat loss or increased heterogeneity of land use (Tscharntke et al., 2012).

Decades of research on productive landscapes have led to various land management proposals. For food production and biodiversity conservation in agricultural landscapes, two approaches are often discussed: integrated landscapes that conserve forests and productive areas on the same land, though with lower yields (land sharing), and increasing sustainable production on existing farmland while keeping conservation areas separate (land sparing) (Tscharntke et al., 2012). A theoretical model for biodiversity conservation in productive landscapes proposes a minimum of 40% forest cover, with one large patch (30%) and smaller patches (10%) (Arroyo-Rodriguez et al., 2020). In South America's Caatinga, ideal sustainable landscapes would have 50% natural dry forest and 50% suitable farmland, emphasizing crop diversification and land restoration (Araujo et al., 2020). Recent evidence has highlighted the imperative need to preserve more natural areas in these productive landscapes to sustain essential ecosystem services and mitigate desertification risks from intensive and unsustainable land use (Araujo et al., 2022). However, the direct impact of these models on biodiversity indicators remains unexplored and unproven.

Birds are a diverse and taxonomically well-recognised group, and are important as seed dispersers, pest controllers and pollinators. For this reason, bird communities are used as ecological indicators. (Morante-filho et al., 2015; Sekercioglu, 2012). Forest-dependent bird species in the tropics are sensitive to changes in habitat amount and spatial configuration, but an opposite pattern is observed for non-forest-dependent species (Morante-Filho et al., 2015). Although the positive relationship between bird diversity and landscape heterogeneity is widely accepted, evidence suggests that this relationship varies across geographic regions and biodiversity dimensions (Fluck et al., 2020).

We assessed the relative influence of habitat amount, landscape heterogeneity and spatial arrangement on the diversity of open-area, generalist and forest-dependent bird species using bird field data from the Caatinga dry ecosystems of northeast Brazil. Firstly, we expect that agricultural landscapes with a greater amount of habitat, which are more heterogeneous and less fragmented, will retain a greater α-diversity and β-diversity of birds. Forest species should be more favourably correlated with forest cover, while birds from open areas should show the opposite patterns. Secondly, we expected that the proportion of different land cover types and the distance between samples would be responsible for maintaining β-diversity of birds in the agricultural landscape (Silva et al., 2017).

Our results showed that the reduction of available natural habitat drastically affected the total bird assemblage and forest specialist bird species. For open-habitat bird species, on the other hand, we found a positive effect of landscape heterogeneity. Therefore, because bird communities are composed of species with different patterns of response to disturbance, our results reinforce the importance of classifying species assemblages on productive landscapes by ecological traits (Morante-Filho et al., 2015; Vargas-Cárdenas et al., 2022). Our results reinforce the idea that changes in landscape structure and matrix landscape determine the dominance or extinction of functional groups, which in turn influence ecosystem functioning, and are consistent with previous evidence from other regions of the world (Morante-Filho et al., 2015; Lecoq et al., 2021; Vargas-Cárdenas et al., 2022).

Two principal contributions of this study emerge: forest loss in anthropogenic dry forest landscapes is the main driver of bird biodiversity loss, both for total bird communities (in terms of α and β) and for forest specialist and generalist bird species. Furthermore, we detected a positive effect of landscape heterogeneity when open-habitat bird species are considered. This is particularly interesting given that the Caatinga's dry forests have been reduced to their original extent. The remaining forests are highly fragmented at the landscape level and are severely affected by chronic anthropogenic disturbance (Silva et al., 2017; Antongiovanni et al., 2020). Habitat loss remains one of the main threats to bird conservation in the Caatinga dry forest, even though the forest is highly fragmented, and the largest fragments are isolated.

Our results support the habitat amount hypothesis, which suggests that habitat amount is a stronger driver of landscape-scale species richness and diversity than spatial arrangement (Fahrig, 2013). These results are consistent with other tropical forest studies showing that even small fragments of secondary vegetation are important elements for maintaining regional biodiversity (Arroyo-Rodríguez et al., 2020). Promoting connectivity between patches could favor bird conservation in agro-landscapes and subsequent ecosystem services, especially in agricultural landscapes such as the Caatinga (Araujo et al., 2020).

Our results highlight the importance of landscape heterogeneity in mitigating biodiversity loss in human-dominated landscapes (Araujo et al., 2020). In the Caatinga, enhancing vegetation connectivity and crop diversity can support various bird functional groups and ecosystem services. The percentage of land cover, especially forest and pasture + crop, is crucial for β-diversity in tropical agricultural drylands. It is important to note that in our study we used forest cover as the main habitat for forest species, but not for open vegetation species. Although the presence of open-area specialist species is important in the Caatinga (Araujo et al., 2012), our results do not reveal that these species are exclusive to a particular habitat, but they can enhance the landscape in a more generalist way. This result may suggest that open habitat species in the Caatinga may be more resilient to disturbance (Araujo et al., 2020).

Natural habitats in productive landscapes are important for the conservation of α and β-diversity (Fahrig, 2013), for this reason, landscape planning in our study region should prioritize natural forest inclusion and restoration at the landscape level (Araujo et al., 2020; Arroyo-Rodríguez et al., 2020). Nonetheless, the importance of heterogeneous production landscapes highlights the importance of mixed approaches to land management. The importance of spatial heterogeneity and edge density for forest birds in this study underscores the role of landscape elements beyond habitat amount for effective land management (Araujo et al., 2012; Hammill et al., 2018; Martin et al., 2020). For example, edge density has been suggested to have detrimental effects on species that inhabit native vegetation, but the magnitude of the effect depends on the specific species' response (Ries et al., 2004). For this reason, the implementation of management strategies that focus on the ecological characteristics of species in agro-landscapes and the identification of specific elements such as the extent and shape of natural edges can be an important part of management approaches.

According to our findings, landscape sustainability in the Caatinga dry forest requires preserving as many natural habitats as possible, but also promoting heterogeneity on current productive lands. Landscape heterogeneity is linked to the qualitative or quantitative variation of landscape elements, influencing various biotic and abiotic processes such as species diversity in different taxa (Hammill et al., 2018; Martin et al., 2020; Tonetti et al., 2023). Our results confirm that landscape heterogeneity increases β-diversity, especially when forest, pasture and crops are present at the landscape scale, but the current trend of land management in the Caatinga dry forest region promotes accelerated desertification processes (Araujo et al., 2020). Promoting landscapes that integrate both land-based management to conserve and restore natural habitats is a priority. Landscape heterogeneity in productive lands has been proposed as a key management strategy in Caatinga landscapes (Araujo et al., 2020). Therefore, we believe that integrative approaches between separate land management (land sparing) and land integration for production and conservation (land sharing) should be considered in the Caatinga, but in a spatially dependent context (Balmford, 2021).

Habitat amount influences α and β-diversity of birds in tropical dryland agricultural landscapes. While forest habitats boost α-diversity for all groups except open-area birds. Additionally, Landscape heterogeneity supports open-area species, while edge density aids forest specialists alongside habitat amount. Furthermore, besides forest cover, cropland ecosystems enhance bird β-diversity, promoting variation and species diversity in agricultural landscapes. Therefore, we propose general strategies to achieve a balance between food production and biodiversity conservation in Caatinga and other drylands: Conservation and restoration of forest habitats by protecting existing forest areas and restoring degraded areas, and by providing economic and political incentives for the conservation of dry forests. Additional strategies are the conservation of forest remnants adjacent to croplands, variability and crop rotation.

All authors contributed to the study's conception and design. Data collection in the field was carried out by Mônica da Costa Lima and Helder F.P. de Araujo, and data analysis and interpretation of the results obtained were carried out by Mônica da Costa Lima, Fredy Alvarado and Helder F.P. de Araujo. First version of the manuscript was carried out by Mônica da Costa Lima under supervision of Fredy Alvarado, and Helder F.P. de Araujo. All authors read and approved the final manuscript.

Data and R code are available in (Lima M. 2023: https://figshare.com/projects/The_importance_of_habitat_amount_and_landscape_heterogeneity_for_bird_diversity_in_tropical_dryland_agrolandscapes/190020